The rash of new information technology has raised major problems for librarians, faced with a flood of new kinds of media, such as audio and video tape or computer disks. Most of these technologies were designed and manufactured without permanence as a prime consideration. A worse problem is the short life of the reading devices; even the less than permanent tapes or disks often do not deteriorate until long after the machines to read them become unavailable. Digital preservation thus depends upon copying, not on the survival of the physical media. Librarians must prepare for reformatting as a regular step in information management. In this, they must join with computer center managers, individual computer owners, and all others who need to keep machine-readable data for more than a few years. The good news is that digital media get steadily cheaper and smaller, so that the problems become steadily easier to solve, and that once something is digitized, copying can preserve the information exactly.
The new media pose several problems to librarians and archivists needing to store them for long periods.
Digital media can be copied without error. Thus survival of digital information does not depend upon the permanence of a particular object, but upon widespread distribution of the information, and regular refreshing of it onto new technology. In general the lifetime of the technology is limited by commercial concerns: as technology improves, old storage technologies disappear. For example, half-inch 9-track magnetic tape densities have gone over the last twenty years from 200 bpi to 556, 800, 1600 and 6250 bpi. Today it is difficult even to find an 800 dpi tape drive, let alone a 200 or 556 bpi drive, or any kind of reader for the 7-track format which preceded 9-track on the same kind of physical tape. And now all half-inch tape equipment is threatened by the more compact DAT (digital audio tape) and 8mm videotape devices.
From the standpoint of a preservationist this example is a particular problem because the new materials are less durable (both DAT and 8mm video are helical-scan tapes, which generally have inferior lifetimes to linear recording tapes). The new systems are changing too: we already have two styles of 8mm videotape. Thus, the only security is copying to new devices regularly.
Reformatting, instead of being a last resort as material physically collapses, will be a common way of life in the digital age. The steady decrease in the cost and size of digital media would make this attractive even if we were not driven to it by technological obsolesence. As new media become widely used in the computer industry, and old media become ever less available, libraries should expect to transfer their digital holdings, either by transcribing files themselves, or (more likely) by buying copies on new media from somebody else. These new copies should be not only in a newer physical format, but in whatever software format is closest to standard at the later time. Peter Lyman (of the University of Southern California Library) points out that while paper deteriorates when used, digital media often not only do not suffer from being used, but benefit from use: if something is used, the librarian will find out faster when the users want it on a new device.
Surprisingly to the novice, the software formats are often longer-lived than the physical devices. Unix word processing software from 1975, for example, is still in use; the storage devices (low density magnetic tape, removable disk packs, and so on) of that time are gone. Unfortunately there is also a much wider variety of logical formats, and much more varied expertise is required to deal with the software content rather with the physical material. The same is true traditionally, of course: a specialist in paper conservation can work on a book regardless of the language it is written in, while a specialist in cataloging one discipline or reading one language normally works only with books which meet that condition. Similarly, it takes fewer experts to keep up with tape and disk machines than it does to keep up with the much wider variety of word processing systems.
On the good side, the intervention of machinery between the actual object and the reader means that the users are unlikely to become emotionally attached to the particular physical media, and thus reformatting of advanced technology should not produce the objections that accompany reformatting of books. It is not expected that computer media will have any artifactual value, except odd examples in the Smithsonian. Also, the steady progress of technology means that the difficulty of keeping any particular item decreases with time. To return to the discussion of punched cards, nobody would want to keep them in that format any more, even if it were practical, since at only 80 bytes per card, 100 pounds of punched cards fit onto one 3.5 inch 3/4-ounce diskette today.
What are the physical formats that are involved? There are many kinds of new format technologies, including:
These devices vary from some which are quite durable (SONY claims a 100-year life for its large WORM cartridges) to others which are fragile (5.25 inch floppy diskettes were often very cheaply made and will not stand hard handling). For those interested in the physical durability of the new media, there is a good review by Saffady. However, it is not usually durability, but rather obsolescence, that poses a threat to the material on the new technologies.
Analog media pose even worse problems. In addition to the technological obsolescence shared with digital media, and which is now wiping out the vinyl record and the devices to play it, for example, the analog media can not be copied without deterioration and are not necessarily very durable. Both audiotape and videotape, for example, are comparatively fragile, especially if used; a book can be read many more times without falling apart than a videotape can be played without major deterioration.
Since analog material degrades with copying, it should be converted to digital at the first practical opportunity, and carried forward in digital form. Digital material is not only copyable without error, but it gets steadily cheaper to store with time. The sooner the librarian converts to digital, thus, the better off we are. Against this is the high cost of conversion, and there is a danger that a low-quality conversion will cause many users to wish to go back to the originals. For example, Michael Ester of the Getty has measured the sensitivity of art historians to different quality digitizations of the same photograph, and demonstrated the important of good quality. Unfortunately, business practices in this area are counter to what libraries would like: some businesses digitize photos only at low resolution not only to save money, but to be sure that no one can use the digitized version (relatively easy to copy) as a substitute for buying the original (somewhat harder to copy illegally).
Converting to digital is now feasible for sound, but digitizing video is still not a routine process which libraries can use. There are also some who think that even with sound, the digital copy is degraded from the analog version (at least enough to keep several hi-fi magazines in business).
Digital storage is made of bits. The storage and preservation of these bits is independent of what they represent. The interpretation of them, on the other hand, is independent of the medium on which they are stored. Thus, conversion and storage breaks down into two parts:
The physical devices, as mentioned, change regularly. Any solution should be viewed as temporary. A library may, however, choose a more or less temporary solution. Compare, for example, WORM optical disk platters with 8mm or 4mm helical-scan tape. The following table shows some basic parameters (SONY 12inch WORM platters are chosen as an example):
Properties of Some Digital Media (1992) WORM 8mm Exabyte DAT Capacity (Gbyte) 6 5 1.3 Cost $340 $5 $9 Wt (oz) 40 3 1.5 Dimensions (inches) 12.75x13.5x.6 4.0x2.6x.75 3x2.6x.6 Volume (cu.in.) 103 5.5 2.4 Long life? yes no no GBytes/cu ft 100 1100 480 GBytes/lb 2.4 27 14 MBytes/$ 17 1000 140Of course, the money spent on the WORM platter gives many benefits. Its life is expected at 100 years, while the tapes are probably only good for a few years. The platter is much more resistant to temperature and humidity variation. It is less likely to be destroyed by a faulty mechanism. But the most important advantage for system designers is the random-access property of the platter, and this makes it more attractive for online use, rather than for archiving. If the only question is preservation, making and distributing multiple copies of the cheaper media is more sensible and protects against other possible problems (e.g. fire or theft). The column ``long life'' does not mean permanent enough to keep forever \- it only reflects the relative lifetime of the devices here, all of which may turn out to have physical lives that exceed their technological life. The library with only a small number of tapes or platters, of course, may find that the cost of the reader also matters \- a reader for the SONY WORM platters costs about $20,000 while DAT and Exabyte drives cost only $1000-$3000 today. This table is somewhat biased in favor of the physically smaller media, since it ignores the need to arrange them in some way that permits access; one can not fill a cubic foot with DAT cartridges and expect to find one that is in the middle.
If the material must be stored on-line, with relatively close to random access, the space and costs are considerably higher. Here are some approximate numbers.
Costs of On-line Storage WORM Magneto-Optical Winchester jukebox jukebox magnetic disks Capacity (GB) 300 40 2.2 Cost $150K $30K $3.7K Size (cu ft) 20 4 .1 GB/cu ft 15 10 20 MB/$ 2 1.3 .6Not surprisingly, these costs are much higher than those for the demountable storage devices. But again, with these media the time required to present any random item is measured in either seconds (for the jukeboxes) or milliseconds (for permanently mounted media), whereas to read through a full Exabyte cartridge may take an hour. Transfer rates are also generally higher on the disk devices. If the primary intent of a library is access, rather than mere preservation, the amount of delay and the need for an operator to mount a tape or disk must be weighed against the cost.
The above tables are only true in 1992. For example, ICI has developed a material they call `digital paper' which offers a promise of even higher density and lower cost, along with long life and non-erasability. It is too soon, however, to say whether this is the likely next device on the market. Currently, the biggest activity is in rewriteable 3.5 inch magneto-optical disks, but the capacities are fairly small in our terms (128 Mbytes).
Unfortunately, just having the bits on some media is not enough. The user needs to be able to find and interpret the bits he or she wants. In some cases, vendors use proprietary software to access data and do not provide details, sometimes as a way to prevent illegal copying. Such techniques further aggravate problems of future access, and require attention to the availability of the software that reads a particular vendor's data, as well as the availability of the device. Standards may help in the future.
The question of format is much more complex. A great many kinds of information are kept in digital forms. No simple table can summarize the choices. Most of the decisions will be made far outside the library community, and because of the economies and practicalities of sticking with what is decided for general industry, the libraries should expect to go along.
On balance, libraries should focus for the moment on text and audio. These are practical and there is some understanding of what should be stored. Images are just about ready for the more advanced organizations to try reformatting. For data and video, libraries should probably stage holding actions with current technology until progress is made by outside standards committees.
Technology for doing conversion of older digital formats to newer formats, either physical or logical, is still quite primitive. The text formatting situation is particularly serious. When faced with a new typesetting tape, for example, it is still sometimes necessary to resort to an octal dump followed by comparison with the printed book. The documentation for the internal formats of many systems is inadequate, either for proprietary reasons or simply because it is not worth doing well if the typesetting system is not widely distributed.
Even if the previous format is known, the converter is normally reduced to writing a specialized program. Sometimes languages like Lex will help, but often there is no really appropriate tool. There has been some work on software to assist in format conversion, for example the Chameleon project at Ohio State [Mamrak 1989] but nothing that is generally accepted. As mentioned before, the problem is that most typesetting and WYSIWYG formatting systems deal only in appearance (e.g. italic, bold) and not in meaning (e.g. cited journal name, section heading). It may or may not be possible to guess at the correct meaning in a particular context, but there are no programs that do so in general.
The situation is somewhat better for graphics where the problem is purely one of format and not of intellectual content. Among the public domain graphics packages are:
Package Author FTP address PBMPLUS Jeff Poskanzer nl.cs.cmu.edu:/pub/pbmplus Fuzzy Pixmap Michael Mauldin nl.cs.cmu.edu:/usr/mlm/ftp/fbm Img-whatnot Paul Raveling expo.lcs.mit.edu:contrib/Img.tar.Z Xim Philip R. Thompson expo.lcs.mit.edu:contrib/xim3c.tar.Z. TIFF Library Sam Leffler ucbvax.berkeley.edu:/pub/tiffIn addition, various commercial packages address the same problems.
The situation for data is even worse than for text; so far, special documentation is required for each file. The only remotely agreed upon structure is relational files, i.e. records with fields, and there is no standard way to write relational databases to an interchange format nor is there likely to be any. The point, of course, is not just to get the numbers out but to have them identified in some way.
Scanning paper is perhaps the one part of conversion which is fairly well understood: the Cornell project on preservation of books by scanning and reprinting, for example, is finding costs of about 10 cents per page to scan a book of 300 pages, plus storage and refreshing costs of under 1 cent per page per year (again for books of 300 pages).
There is great economy of scale in maintaining new media \- not so much the devices, but in sharing the expertise. Centralized archives exist today \- for example at Oxford University for literary texts, or the University of Michigan for social science data. Cooperation is essential for libraries to manage. Not only will standard formats help a great deal, but copying digital information is much easier than generating it. Particularly where lots of intellectual effort is required (e.g. indexing pictures or audio) it makes no sense for different groups to work on the same file. It will be much easier for libraries to obtain updated versions from central sources than for each library to manage all its own reformatting. A cooperative program for converting digital information would be appropriate. It should involve both maintaining records of who has what, seeing that conversions are done only once, and seeing to it that important materials are kept in more than one place.
There are a great many existing archives. In the Appendix is a short list, several years old, of machine-readable archives within Canada, the UK and the US which was collected collected by Prof. Robert Kraft of the University of Pennsylvania. It is shown merely as an example of the kinds of centralized archives which already exist; lists of such archives can be found both online and in conventional publications (Neumann 1991, Lancashire 1991). There are already secondary services which collect knowledge about machine-readable texts, such as the Center for Electronic Texts in the Humanities (Rutgers and Princeton), whose catalog is in turn online under RLIN.
Efforts should be made to find funding to put archives of this sort on a firm footing. In the long run, it will be cheaper to have sites of responsibility for particular materials than to attempt to duplicate expertise in specialized records conversion. Furthermore, sites of this form will function in the traditional way to link scholars working in the same area, even if these links are now via electronic mail rather than by physical travel to the site.
Most of the archives above are textually oriented. Support is needed to extend computer archives in the multi-media areas, and also in the areas of numerical data and programs. Some attention is also needed to ways of indexing the material in these archives, as there is no straightforward way to locate information, particularly non-textual information, in machine-readable form. Bibliographic guides will have to be developed, and even the format these should take is not clear.
Industrial and government computing groups have often had the problem of maintaining old tape files, and have stimulated much of the work in records management for machine-readable files. Unfortunately the history here is often poor: it is said that we know more about the 1860 census than the 1960 census, which was written on a very early form of computer tape and which no one realized in time should be copied to newer media.
Copyright rules may pose a severe problem. The copyright life of much material greatly exceeds the technological life of the media it is recorded on. A library may well be able to preserve something only by copying it, but may have difficulty obtaining permission to do that. With luck the original issuer will publish the material in a new format, but as the devotees of old records and movies can testify, this does not always happen. The concept of ``fair use'' should be extended to include the copying to new media of a publication which is no longer available for sale, and the distribution of the copy to other libraries. At the moment, there is an exemption for copying of a publication that a library already owns, but not for distribution to libraries that do not own the original being copied. In the next century, as the United States changes to a ``life plus 50'' standard for copyright expiration, it may even be hard to determine whether or not a work is still copyrighted.
The law may also pose problems even for information where the copyright law does not bar conversion. A library might wish to store information derived from old, copyright-expired publications; or from public domain sources; or from publications of the university to which it belongs. However, it is unclear whether a copyright can be asserted in a straightforward transcription to machine-readable form of a non-copyrighted document or audiotape. The rule that a copyright can only be claimed for the result of creative activity has recently been interpreted to say that (a) collecting and alphabetizing names into a telephone book is not copyrightable (Feist Publications v. Rural Telephone Service) but (b) numbering the pages in a Government-provided text is copyrightable (West Publishing v. Mead Data Central).
Although measuring the creativity in these publications is like arguing whether a concrete block is more or less intelligent than a 2x4, the legal uncertainty causes all vendors of electronic information to require specific contracts limiting the use of the purchased data, imposing administrative burdens on everyone.
Libraries, depending on the circumstances, might be on either side of this issue. Presumably, the ability to enforce a copyright in a converted text will encourage someone to do a conversion. On the other hand, if the issuer then disappears, it becomes difficult to get permission to do the next conversion when that becomes necessary. Again, it would be useful to have a rule that the copyright lasts only as long as someone is willing to sell the protected item.
The copyright law also affects the question of multiple copies. The United States is developing a very high speed digital network which will soon provide 45 Mbits/second between major sites and should be at 1 Gbit/second before the end of the decade. Thus, libraries may well feel that there is no need to have their own copy of something, since the delay it getting it remotely may be negligible. But, depending on copyright rules and the developing business practices in the electronic information industry, it may be difficult to arrange access permission. On the other side, it would be unfortunate if the existence of the network meant that important documents were kept in only one copy which was at risk of disappearing (e.g. residing in some individual's personal disk files, rather than being kept in some permanent institution's records management system). Finally, the insertion of copy-protection into sound recordings acts against the interests of libraries and long-term permanence of the original material. Deposit of an unprotected copy in some trustworthy institution is probably the best solution, but not something the industry is pursuing.
In this new world, preservation means copying, not physical preservation. The Commission on Preservation and Access is already involved in this basic idea through its efforts in reformatting. It should attempt to encourage librarians to cooperate in learning how to deal with preservation in the new technological context. What steps can be taken?
Most important, the Commission needs to inspire a new view among librarians. Preservation means something different with digital media: it means copying, not physical preservation, and it requires more attention to long-term costs (some analog media are also best preserved by copying, of course). There are analogies to the old problems in the new era, but not exact ones. For example, rebinding is irrelevant with digital media, where no one is attached to the physical form of the device; instead we have refreshing, or copying. Theft is still a problem, but it is more likely to be in the form of illegal copying rather than stealing a physical device. Mis-shelving of books is gone, but computer viruses and accidental or deliberate erasing (e.g. to gain an advantage in a competitive undergraduate course) will replace it. Systematic problems such as acidic paper correspond to the unfortunately much more rapid bankruptcy of suppliers. Librarians may not warm to an analogy with someone who buys new clothes because the styles have changed rather than because the old ones have worn out, but that's the way computer devices develop.
The preservation task for new materials does share with microfilming a need for collaboration. A centralized clearinghouse of some sort is needed to avoid doing the same conversion twice; and it is likely that every library will not wish to be involved in creating digital material in new formats, most merely using it, and creating it through consortia or contract houses.
In particular, cooperation on these issues is needed not only among librarians, but with computer centers and others involved in records management. The problems facing the libraries are also faced by every computer operator, and efforts should be made to share information between all affected groups.
Of the technical issues, the most important is to develop better conversion technology. It is extremely frustrating today to encounter unfamiliar text in machine-readable form: often considerable human expertise is needed to extract what the user wants from it. Research in ways of converting formats or adapting old software would be particularly important.
What will this cost? Let us try to imagine the librarian of a few years from now buying a book. First, consider a model just like the library of today \- shelves full of stuff, readers sitting at tables \- except that the shelves are full of tape cartridges and there are computer screens on the tables in front of the readers.
Suppose that a book is purchased as a batch of images on tape from the vendor; assume 300 pages at 100 Kbytes per page, or 30 Mbytes total. Perhaps it is purchased on a DAT cartridge; 30 books will fit on the one cartridge, whose material cost might be $10. This sounds like the book will only cost 35 cents, but remember that the paper, ink, cloth, glue and postage for a $25 book does not cost much more than $1 today. So the book will probably still cost $25; the publisher still has all the other costs.
However, every few years the library will have to buy a new device and copy the tape. Copying such a tape takes about 30 minutes, or 1 minute per book. The major part of the cost will be the new tape (suppose another 70 cents) and the cost of somebody to watch the drive (at $10/hr, perhaps 20 cents). Thus, a superficial estimate suggests that it will cost $1 every 3 years, or 30 cents per year, to refresh the book. In addition to the $20 to buy the book, therefore, the library would have to set aside enough money to generate 30 cents per year, or perhaps another $6. More thorough estimates at Cornell University suggest a cost of $1.37 per book per year for refreshing.
This may sound like a disadvantage; but remember that the single cartridge which holds 30 books fits in half the space of an ordinary book. Assume that shelf space normally costs in the neighborhood of $1 per year per book,* and that half of that is the kind of building cost which varies proportionally to square feet; then, this is reduced from $0.50 to $0.01 for the cartridges.
This number comes from Bob Hayes, but to justify it, consider that in a square foot of floor space a library can put about 80 books \- 1 book per inch, shelves about 12 inches deep with support, and six to seven shelves high. Assuming that the shelving requires a square foot of aisles for each square foot of shelves, this means about 40 books per square foot. Renting commercial space with heat, power, etc. is about $40 per square foot per year.
This saves almost 50 cents per year per book, and will compensate for much or all of the cost of copying, provided that the accounting system of the university or other organization to which the library belongs can actually equate the capital cost of providing shelf space to the running cost of performing digital copies.
If the books can be bought as typesetting tapes instead of images they will occupy 1/20 or so the digital space, and thus the total costs of the computer media and operations will decline, leaving the library somewhat better off. Most importantly, it is more likely that a library will be able to bulk purchase a large number of books from some archive or publisher cooperative at a reasonable cost, and expect them to maintain the material as formats change.
It might be thought that libraries will be able to get revised formats of books from the original publisher. However, it is not clear how the pricing will work. At least today, no publisher sells a second edition of a book to a library at a reduced cost if it owns the first edition. Furthermore, the vast majority of books come out in only one edition, and it is not clear that will change.
With time, this process will at least be getting cheaper. As storage densities increase, whatever information is being copied today will fit in less space tomorrow (for example, a single Exabyte cartridge will hold the equivalent of 95 reels of 800-bpi magnetic tape, and the cartridge will fit in 6 cubic inches while the 95 reels of tape need more than 6000 cubic inches).
The other obvious model is that the library does not buy most books, but obtains them on-demand from some centralized archive. In this model, the library still contains the tables, and has some machines on them, but the book content is not obtained from any kind of local stack, but from a remote link to a centralized archive. Users do searches in secondary services or in catalogs, and then retrieve papers or books from the archive servers. What will the data transmission cost? At the moment, most Internet sites pay a bulk once-a-year price; there is no charge for individual messages. What might a charge be? At present they are still rather high: for example, A DS-1 (1.5 Mbit) line across the U. S. costs about $600 per hour. That means that $1 will move about 1.2 Mbytes, so a 30 MByte book would cost $25. But the NREN is likely to be much cheaper, at least if present political and technical trends continue. Just as today, postage is not a major part of the cost of getting books onto the shelves, in the future, data transmission is not going to be most of the cost of getting bits into the library. In fact, the costs of data tranmission are likely to be much less than the savings of not having to locally catalog and manage the books.
The major issue will be copyright permission. It may be technically feasible to move bits even today, but administratively it may be a horror. Work is needed to sort out the legal and practical problems of making file-sharing possible. Some researchers (e.g. Brian Kahin of Harvard, John Garrett of CNRI, and Marvin Sirbu of CMU) have already been working in this area. What is needed most are some prototype projects, to find suitable business models and see which of the problems envisaged are real and which imaginary, and sketch out a suitable balance by which the users, libraries, publishers/authors, and information industry vendors share in the advantages and the costs of the new technology as applied to books. The computer software industry, for example, seems to be converging on site licenses or concurrent-use licenses as a reasonable pricing mechanism; perhaps this will serve as a guide to the libraries.
The message for the librarian in the digital world is that
The comments of Clifford Lynch, Stuart Lynn, and Peter Lyman have been very helpful in preparing this article.
McGill (Univ): Kierkegaard-Wittgenstein Project
Montreal (Univ): Instut d'Etudes Médiévales
Montreal (Univ Quebec): Centre d'Analyse de Textes par Ordinateur
Newfoundland (Univ): Folklore & Language Archive
Ottawa (Dept Communications): CHIN = Canadian Heritage Information Network
Ottawa (Carleton Univ): Centre for Editing Early Canadian Texts
Quebec (Laval Univ): Bibliographical Information Bank in Patristics
Toronto (Univ): RIM = Royal Inscriptions of Mesopotamia Project
Toronto (Univ): REED = Records of Early English Drama
Toronto (Univ): CCH = Centre for Computing in the Humanities
Toronto (Univ): Dictionary of Canadian Biography
Toronto (Univ): DOE = Dictionary of Old English
Toronto (Univ): DEEDS Project = Documents of Essex, England Data Set
Toronto (Univ): Greek Index Project (Pontifical Institute of Medieval Studies)
Waterloo (Univ): Centre for the New Oxford English Dictionary
Vancouver (Univ British Columbia): UBC Data Library
Vancouver (Simon Fraser Univ): RDL = SFU Research Data Library
Cambridge (Univ): LCC = Linguistic Computing Centre
Edinburgh (Univ): EUDL = Edinburgh University Data Library
Edinburgh (Univ): Greek Text Database
Essex (Univ): ESRC = Economic and Social Research Council Data Archive
Glasgow (Univ): DISH = Design and Implementation of Software in History Project
Lancaster (Univ): UCREL = Unit for Computer Research on the English Language
Leeds (Univ): Centre for Computer Analysis of Language and Speech
London (Univ): School of Oriental and African Studies
London (Univ College): Survey of English Usage
Oxford (Press): OED = Oxford English Dictionary
Oxford (Press): Oxford Shakespeare
Oxford (Univ): OTA = Oxford Text Archive
Oxford (Univ): Lexicon of Greek Personal Names (Bodleian Library)
Southampton (Univ): AIE = Archaeological Information Exchange
York (Univ): Graveyard Database
AZ Tucson (Museum): Documentary Relations of the Southwest
CA Berkeley (Univ CA): Anthologies of Italian Music and Lyric Poetry of the Renaissance
CA Berkeley (State??): SDB = State Data Bank [supplements ICPSR/Michigan]
CA Berkeley (Univ CA): Sino-Tibetan Etymological Dictionary and Thesaurus
CA Davis (Univ CA): Project Rhetor
CA Irvine (Univ CA): TLG = Thesaurus Linguae Grecae
CA Los Altos (Inst): PHI = Packard Humanities Institute
CA Los Angeles (UCLA): Computerization of Arabic Biographical Dictionaries for the Onomasticon Arabicum
CA Malibu (Udena Publ): CAM = Computer-Aided Analysis of Mesopatamian Materials
CA Menlo Park (Cent): Center for Computer Assisted Research in the Humanities
CA Riverside (Univ CA): Laboratory for Historical Research
CA Riverside (Univ CA): Biographical Data Base for the Soviet Bureaucracy
CA San Diego (Univ CA): International Electronic Archive of the Romancero
CA Santa Barbara (Univ CA): Domesday Book Database
CA Santa Monica (Getty Art Hist Info Prog): Provenance Index
CA Stanford (Univ): Institute of Basic German
CO Colorado (Univ): Siouan Languages Archive
CO Boulder (Univ CO): CCRH = Center for Computer Research in Humanities
CT Hamden (??): Encyclopedic Thematic Catalog of Russian Sacred Choral Music
DC Washington (Georgetown Univ): Electronic Text Repository
DE Newark (Univ DE): Massachusetts Tax Valuation List of 1771
FL Talahasee (FL State Univ): Center for Music Research
HI Manoa (Univ HI): Salish Lexicography
IL Chicago (Univ): ARTFL = American and French Research on the Treasury of the French Language
IL Chicago (Newberry Lib): County Boundaries of Selected United States Territories/States
IL DeKalb (N IL Univ): Tai Dam Dictionary and Text on Computer
IL Urbana (Univ IL): Hymn Tune Index
MA Boston (Center??): Census of Gothic Sculpture in America
MA Cambridge (Harvard Univ): Boston Dainas Project
MA Cambridge (Harvard Univ, Boston Univ): Perseus Project
MA Williamstown (Getty Art Hist Info Prog): Art and Architecture Thesaurus
MD Baltimore (Johns Hopkins Univ): CAL = Comprehensive Aramaic Lexicon
MI AnnArbor (Univ MI): Family Life and Conditions in the US, 1888-1936
MI AnnArbor (Univ MI): ICPSR = Inter-university Consortium for Political and Social Research
MI Dearborn (Univ MI): Comprehensive Computer Data Bank of the Medicinal Plants of Native America
MS Hattiesburg (S MS Univ): Faulkner Computer Concordance
NC Chapel Hill (UNC): DBAGI = Data Bank for Ancient Greek Inscriptions from Athens
NC Durham (Duke Univ): DHDB = Duke Humanities Data Base
NC Durham (Duke Univ): DDBDP = Duke Data Bank of Documentary Papyri
NC Winston-Salem (Museum): Index of Early Southern Artists and Artisans
NH Hanover (Dartmouth Univ, Princeton Univ): Dartmouth Dante Project
NJ New Brunswick (Rutgers Univ): Medieval and Early Modern Data Bank
NJ New Brunswick (Rutgers): Lexicon Iconographicum Mythologiae Classicae
NJ Princeton (Inst Advanced Studies): Greek Inscriptions from Asia Minor
NJ Princeton (Univ): American Founding Fathers Project
NY Binghamton (SUNY): Italian Madrigal and Related Reportories: Indexes to Printed Collections, 1500-1600
NY Buffalo (SUNY): WNY-ARCH = Western New York Archaeology
NY Ithaca (Cornell Univ): Greek Inscriptions from Attica
NY Ithaca (Cornell Univ): Cornell Blake Concordance Texts
NY New York (Columbia Univ): Women in Religious Communities: Italy 500-1500
NY New York (Columbia Univ): Data Archive, Center for Social Sciences
NY New York (Columbia Univ): Data Base on Labor Unrest in Imperial Russia
NY New York (Columbia Univ): Great Dictionary of the Yiddish Language
NY New York (Columbia Univ): Buddhist Canon Project
NY New York (Jewish Theological Seminary): Talmud Text Databank
NY New York (NYU): The Verdi Archive
OH Cleveland (Cleveland State Univ): Century-of-Prose Corpus
PA Philadelphia (Drexel Univ): The Latin Writings of Milton
PA Philadelphia (Univ PA): CCAT = Center for Computer Analysis of Texts
PA Philadelphia (Univ PA): Language Analysis Project
RI Providence (Brown Univ): WWP = Women Writers Project
RI Providence (Brown Univ): Romanian Love Incantations
RI Providence (Brown) ??: Nelson Francis Brown Corpus
TX Dallas (Theol Seminary): Biblical Data Bank (CD-ROM)
TX Del Valle (??): Chol (Mayan) Dictionary Database
TX Edinburg (Pan American Univ): RGFA + Rio Grande Folklore Archive
UT Provo (Brigham Young Univ): HRC = Humanities Research Center
UT Salt Lake City (Church of Latter Day Saints): Genealogical Data
WI Madison (Univ WI): DOSL = Dictionary of the Old Spanish Language
WI Milwaukee (Marquette Univ): Works of Karl Rahner Project
Saffady 1989. William Saffady, ``Stability, Care and Handling of Microforms, Magnetic Media and Optical Disks,'' Library Technology Reports, vol. 27, p. 5-117, (January-February 1991).
Mamrak 1989. S. A. Mamrak, M. S. Kaelbling, C. K. Nicholas and M. Share, ``Chameleon: a system for solving the data-translation problem,'' IEEE Transactions on Software Engineering, vol. 15, no. 9, pp. 1090-1098 (1989).
Neumann 1991. Toward Improvement in Electronic Versions of Humanities Texts,'' by Michael Neumann, Computers and the Humanities, v. 25, p 363-375 (1991)
Lancashire 1991. The Humanities Computing Yearbook, 1989-1990, edited by Ian Lancashire and Willard McCarty, Oxford University Press, 1991.